Stephen Greer
@stepheng29.bsky.social
Industry Consultant at SAS Institute. Interested in the tech industry, AI/ML, data science, fintech. 49ers fan.
OpenAI o3 models achieve 87.5% on the ARC-AGI Eval, absolutely blowing away any of the previous models.
December 22, 2024 at 12:47 AM
OpenAI o3 models achieve 87.5% on the ARC-AGI Eval, absolutely blowing away any of the previous models.
YC's request for startups Winter 2025
If you were to start an economy today from scratch....
If you were to start an economy today from scratch....
December 10, 2024 at 5:00 PM
YC's request for startups Winter 2025
If you were to start an economy today from scratch....
If you were to start an economy today from scratch....
Apparently LLMs perform much better in complex tasks when they are told to imitate rather than purely reason. 🎭
🔑 Acting-based prompting:
Treat LLMs as performers, prompts as scripts.
🔑 Acting-based prompting:
Treat LLMs as performers, prompts as scripts.
December 3, 2024 at 1:41 PM
Apparently LLMs perform much better in complex tasks when they are told to imitate rather than purely reason. 🎭
🔑 Acting-based prompting:
Treat LLMs as performers, prompts as scripts.
🔑 Acting-based prompting:
Treat LLMs as performers, prompts as scripts.
AI Agents Stack, Nov 2024
November 30, 2024 at 3:45 PM
AI Agents Stack, Nov 2024
Perplexity keeps shipping very interesting features
November 26, 2024 at 4:52 PM
Perplexity keeps shipping very interesting features
DynaSaur: The AI agent that evolves by coding
→ Writes new Python functions when needed
→ Reuses & optimizes actions via embeddings
→ Builds a growing library of tools
Results:
🏆 Tops GAIA leaderboard
📈 +81.59% w/ human + AI tools
🚀 Excels in complex tasks (level 2 and 3)
→ Writes new Python functions when needed
→ Reuses & optimizes actions via embeddings
→ Builds a growing library of tools
Results:
🏆 Tops GAIA leaderboard
📈 +81.59% w/ human + AI tools
🚀 Excels in complex tasks (level 2 and 3)
November 25, 2024 at 1:50 PM
DynaSaur: The AI agent that evolves by coding
→ Writes new Python functions when needed
→ Reuses & optimizes actions via embeddings
→ Builds a growing library of tools
Results:
🏆 Tops GAIA leaderboard
📈 +81.59% w/ human + AI tools
🚀 Excels in complex tasks (level 2 and 3)
→ Writes new Python functions when needed
→ Reuses & optimizes actions via embeddings
→ Builds a growing library of tools
Results:
🏆 Tops GAIA leaderboard
📈 +81.59% w/ human + AI tools
🚀 Excels in complex tasks (level 2 and 3)
Drag & Drop prompt engineering IDE to eval LLMs for agentic workflow.
All OS and No-code.
All OS and No-code.
November 21, 2024 at 5:27 PM
Drag & Drop prompt engineering IDE to eval LLMs for agentic workflow.
All OS and No-code.
All OS and No-code.
An analysis by Stanford economists estimating that nearly a quarter of all U.S. innovation since 1976 can be credited to highly skilled, foreign-born individuals.
November 21, 2024 at 2:20 PM
An analysis by Stanford economists estimating that nearly a quarter of all U.S. innovation since 1976 can be credited to highly skilled, foreign-born individuals.
From “Barbary Coast,” written in 1933 about San Francisco in 1849.
November 20, 2024 at 2:39 PM
From “Barbary Coast,” written in 1933 about San Francisco in 1849.
AWS just released a new Multi-Agent AI framework
It lets you manage multiple AI agents, dynamically route LLM queries, maintain context across AI Agents and can be deployed locally on your computer.
100% opensource.
It lets you manage multiple AI agents, dynamically route LLM queries, maintain context across AI Agents and can be deployed locally on your computer.
100% opensource.
November 19, 2024 at 3:29 PM
AWS just released a new Multi-Agent AI framework
It lets you manage multiple AI agents, dynamically route LLM queries, maintain context across AI Agents and can be deployed locally on your computer.
100% opensource.
It lets you manage multiple AI agents, dynamically route LLM queries, maintain context across AI Agents and can be deployed locally on your computer.
100% opensource.